K-Means Based SVD for Multiband Satellite Image Classification Assad H. Thary Al-Ghrairi Dr. Mohammed S. Mahdi Al-Taei

Transcription

1 K-Means Based SVD for Multiband Satellite Image Classification Assad H. Thary Al-Ghrairi Dr. Mohammed S. Mahdi Al-Taei 816 Abstract The motivation we address in this paper is to classify satellite image using the singular value decomposition (SVD) technique, the proposed method is consisted of two phases; the enrollment and classification. The enrollment phase aims to extract the image classes to be stored in dataset as a training data. Since the SVD method is supervised method, it cannot enroll the intended dataset, instead, the moment based K-means was used to build the dataset. The enrollment phase showed that the image contains five distinct classes, they are; water, vegetation, residential without vegetation, residential with vegetation, and open land. The classification phase consisted of multi stages; image composition, image transform, image partitioning, feature extraction, and then image classification. The classification method used the dataset to estimate the classification feature SVD and compute the similarity measure for each block in the image. The results assessment was carried out by comparing the results with a reference classified image achieved by Iraqi Geological Surveying Corporation (GSC). The comparison process is done pixel by pixel for whole the considered image and computing some evaluation measurements. It was found that the classification method was high quality performed and the results showed acceptable classification scores of about 70.64%, and it is possible to be approaches % when considering both classes: residential without vegetation and residential with vegetation as one class for SVD method. Keywords Satellite image, Image classification, segmentation, SIC, KMeans, SVD, Moment. 1. INTRODUCTION Remote sensing uses satellite imagery technology to sense the data. Mayank [4] provide a feed-forward neural network for the landcover of Earth. At the early of 21st century, satellite satellite image segmentation, which provides a way to solve the imagery became widely available with affordable [1]. Satellite problem of parametric-dependence involved in statistical image classification is the most significant technique used in approaches using a robust, fault-tolerant, feed-forward neural remote sensing for the computerized study and pattern network. Márcio and et al., [8] presented a methodology for the recognition of satellite information, which is based on diversity landcover classification of satellite images based on clustering structures of the image that involving rigorous validation of the of the Kohonen s self-organizing map (SOM). Sathya and training samples depending on the used classification algorithm Malathi [9] established a segmentation and classification of [2]. It is an extreme part of remote sensing that depends remote sensing images. This classified image is given to originally on the image resolution, which is the most important KMeans algorithm and Back Propagation algorithm of ANN to quality factor in images [3]. Image Classification or calculate the density count, the excremental result found that K- segmentation is a partitioning of an image into sections or means algorithm gives very high accuracy, but it is useful for regions. These regions may be later associated with ground single database at a time. Rowayda [10] Performed an cover type or land use, but the segmentation process simply experimental survey for the SVD as an efficient transform in gives generic labels (region 1, region 2, etc.) to each region. The image processing applications, some contributions that were regions consist of groupings of multispectral or hyperspectral originated from SVD properties analysis in different image image pixels that have similar data feature values. These data processing are proposed. Bjorn Y. [11] established method for feature values may be the multispectral or hyperspectral data the classification of satellite images into multiple predefined values and/or they may be derived features such as band ratios landcover classes. Ankayarkanni and Ezil [12] proposed an or textural features [4]. The powerful of such algorithms is efficient image classification technique for satellite images with depends on the way of extracting the information from huge the aid of KFCM and artificial neural network (NN). number of data found in images. Then, according to these Ankayarkanni B. and et al. Harikrishnan and Poongodi S. [13] information, pixels are grouping into meaningful classes that implemented a cellular with fuzzy rules for classifying the enable to interpret, mining, and studying various types of satellite image and analyzed the quality of classified image. regions that included in the image [3]. Many applications based Akkacha and et al [14] proposed a combination of three on using Landsat imagery in a quantitative fashion require classification methods which are K-means, LVQ (linear vector classification of image pixels into a number of relevant quantization) and SVM (support vector machine). Thwe and et categories or distinguishable classes [5]. These applications use al [15] proposed a method for area classification of Landsat7 image classification as an important tool used to identify and satellite image. The paper is structured as follows. Section 2 detect most relevant information in satellite images [6]. Bin and describes the contribution of work. In Section 3 describes et.al [7] provide a neural network-based cloud classification theoretical frame work. In Section 4 describe the proposed SIC. using the wavelet transforms (WT) and singular value Finally, section 5 describes the results and discussion. decomposition (SVD) to extract the salient textural feature of

2 CONTRIBUTION The proposed method is based on the use of k-means based singular value decomposition (SVD) for satellite image classification. The use of such method enables to study the concepts that concerned with training phase. The study of training phase capabilities leads to improve the classification results. SVD is stand for supervised method depending on predefined dataset stored in the dictionary that firstly established using the k-means. The optimal run of training phase leads to create optimal dataset stored in the dictionary and then used to determine intended classification results when the classification phase is running. Implies, the optimal choice of the dataset indicates an optimal classification results. 3. THEATRICAL BASIS Classification of satellite images can be achieved by unsupervised or supervised procedures, it is performed when the image needs to be assigned into a predefined classes based on a number of observed attributes related to that image. This refers to the task of extracting information from satellite image; the information is assigned into classes according to specific features that distributed in the image [16]. The following sections introduce the concepts of the used features: singular value decomposition (SVD) and moment besides the clustering based on K-Means. 3.1 Singular Value Decomposition Singular Value Decomposition (SVD) is a mathematical tool widely in image classification; it is useful factorizations method in linear algebra [17]. SVD technique is based on a theorem of linear algebra that mentions; a rectangular m n contents of an image or image segment, such that it is used to matrix A having m rows and n columns in which m n, is can distinguish different image segment from each other. Also, it is be broken down into the product of three matrices, as given in used to describe details of small areas found in the image, Equation (1) [18]. which is a useful for image classification [20]. A= (1) Where U is a m n matrix of the orthonormal eigenvectors of AA T called the left singular vectors of A satisfy equation (2), V T is the transpose of a n n matrix containing the orthonormal eigenvectors of called the right singular vectors of A satisfy equation (3), and are the identity matrices of size n and p, respectively, and S is a n n diagonal matrix with nonnegative diagonal entries of the singular values which are the square roots of the eigenvalues of and called the singular values of A, which given in eq.(4) [19], as follows: S= U=... (2) V=... (3) Where 1 2 p, p = min {m, n}, and U = V =. In such case, the SVD value is the minimum value of diagonal terms, which can be used as a distinguished feature for any image segment. 3.2 K-Means Based Clustering K-means is one of the effective unsupervised learning methods that solve the clustering problem. The application of this algorithm on digital image requires being starts with some clusters of pixels in the feature space, each of them defined by its center. The first step is randomly choosing a predefined number of clusters. Second step is allocating each pixel to the nearest cluster. While, the third step is computing new centers with new clusters. These three steps are repeated until convergence. Therefore, the k-means algorithm adopts the following three steps till reaching the final state [9]. 1. Determine the centroid coordinate. 2. Determine the distance of each object to the centroid. 3. Group the object based on minimum distance. 3.3 Moment Based Classification The concept of moment is derived from Archimedes' discovery of the operating principle of the lever. In the lever one applies a force, in his day most often human muscle, to an arm, a beam of some sort. Archimedes noted that the amount of force applied to the object (i.e., moment) is defined as (5) Where F is the applied force, and r is the distance from the applied force to object and s is the order of the moment. In image processing, the pixel value represents the force F, whereas r is the distance between the pixel and the center of the moment. The moment gives an actual indication about the 4. PROPOSED CLASSIFICATION METHOD The generic structure of the proposed method for satellite image classification using K-means based SVD is shown in Figure (1). It is shown that the proposed method is designed to be consisted of two phases: enrollment and classification. The enrollment phase goes to collect the training dataset (referred as A), which an offline phase is responsible on collecting sample image classes to be stored in dataset matrix to be a comparable models. Whereas the classification phase is an online phase responsible on verifying the contents of the test image in comparison with the trained models found in the dataset, this phase depends on the dataset created by the enrollment phase. Both phases are composed of the three preprocessing stages include: image composition, image transform and preparing. Then, the enrollment includes sequenced stages of image partitioning, feature extraction and then clustering to establish the dataset. On the other hand, the classification phase consist of sequenced stages aims to extract the classification features from the employed image unit. In addition, there are an intermediate stages included in the classification are used to achieve the intended purpose are shown in Figure (1) and described in the following sections.

3 818 Enrollment Start Load Multibands Satellite Image Image Composition Image Transform Image Preparation Classification Conditions Setting Phase Classification band. The three bands of greatest value of D are chosen to be combined with each other to make the composed image F R,G,B (i, j) employed in the following stages.... (6) Where, and are the mean and variance of k th band of satellite image of W H resolution as given in equations (7 and 8). Such that, the green band F G, red band F R, and blue band F B are assumed to be the first three bands that possess maximum dispersion coefficient D k as given in equations (9-11): (7) (8)... (9)... (10)... (11) Uniform Image Partitioning Moment (M) Computation Clusters Number (N c) Determination K-Means Clustering Dataset Creation Dataset Storage 4.1 Image Composition 4.2 Image Transform The three estimated bands F R, F G, and F B are converted into newly bands according to YIQ color transformation system. Quadtree Partitioning The Y represents the intensity band, whereas both I and Q represent the chrominance bands. Just the Y band is useful in the present work, which can be noted as F T and estimated according Get New Image Block to the following relation: SVD Computation F T (i,j)= F R (i,j)+870 F G (i,j) f B (i,j)... (12) 4.3 Image Preparation Similarity Measurement This stage is regarded to increase the contrast of the given Image Block Classification material image. Contrast stretching is used to enhance the End End Blocks Figure (1) Block diagram of the proposed SIC method. Satellite image is usually taken in multibands; this stage is aiming to compose the most informatic three bands in one color image given in RGB color space. The dispersion coefficient (D) of the whole image f (i,j) that given in equation (6) is used as a measure for quantifying whether a set of observed details are clustered or dispersed compared to a standard case. This parameter indicates the amount of the information found in each Yes No Classified Image Display appearance of image details, which can be achieved by adopting the linear fitting applied on the input image F T for achieving the output image F P as given in the following equation: F P =af T +b (13) Where, a and b are the linear fitting coefficients given in the following equations, in which Min 1 and Max 1 are the minimum and maximum values of pixels found in transformed image, whereas Min 2 and Max 2 are the intended values of the minimum and maximum of output image pixels. 4.4 Classification Conditions Setting (14) (15) In this stage, the intended conditions of classification status are determined. This conditions are used in both enrollment and classification phases. For the partitioning stage, the maximum block size (B Max ) and minimum block size (B Min ) are set at the situation that gave best classification

4 819 results. This depends on the number of try making for achieving best results. 4.5 Enrollment Phase The enrollment of dataset is an important step in the image classification. It is used for determining the image classes depending on sequenced stages. It is intended to uniformly partition the prepared image (F P ) into equal blocks of size B Max. The reason of using B Max is to make the dataset containing greater number of information related to each class, and make the moment is the feature that recognizes each part. K-Means algorithm is used for grouping these features and then determining the best clusters (centroids) within the resulted features. The image part belongs or closes to each centroid are stored in dataset array to be used in the classification phase. This dataset can resized and scaled down to be half or quarter B Max as needed in the classification. The average of the two successive elements gave new value in the half scaled down dataset, and another averaging leads to get quarter scaled down for the dataset. The Moment is a specific quantitative measure used to represent the information found in each image block. The shape of a set of pixels is a distribution of mass, which can be described by first-ordered moment given in equation (5), where the applied force (F P ) represented the pixel of block and r is the distance from the applied force to the center of block. In such case, the pixel value (F P ) is regarded as the meant force, while the distance (D s ) is determined depends on the position of each pixel (In first, second, third, or fourth quarters) as shown in Figure (2). The moment of each block can be determined as shown below: 1. Compute the Euclidean distance D s between each pixel of a specific block and the center of that block (the difference between the pixel and the center of block) as follows: a. If the pixel F P (i, j) falls in the First quarter then the D s1 is computed by using the following relation: D s1 = (16) M = (21) Where B h is the height of block and B w is the width of block, M (i, j) represent the moment of pixel in a specific block of image, and F P (i,j) represent the pixel value of a specific block at position (i,j), i, and j are indices of the pixel in block of image. DS1 0, 1 1,0 1, 1 (a) Computation of the distance D s1. 0, 2 1,2 i o, j Ds2 o i 0, 0 j i o, j o (b) Compute the distance D s2. 0,3 1, 2 b. If the pixel F P (i, j) falls in the Second quarter then the D s2 is computed by using the following relation: D s2 = (17) i o, j o c. If the pixel F P (i, j) falls in the Third quarter then the D s3 is computed by using the following relation: D s3 = (18) 2, 0 2, 1 Ds3 d. If the pixel F P (i, j) falls in the Fourth quarter then the D s4 is computed by using the following relation: D s4 = (19) Where, are represent the indices of the center block. 2. Compute the moment M p (i, j) of each pixel in a specific block of image by using the following relations: M p (i, j) = F P (i, j) x D s (20) 3. Compute the moment of a specific block (M) in image by using the following relation: 3,0 3, 1 (c) Compute the distance D s3.

5 Calculate the number of pixels N in the image that fall within the range of 2 in the image distribution. 4. Compute the percent (P) of the pixels number (N) in 2 expansion and the number of pixels in whole image (N T ) by using the following relation: (25) 5. The number of classes (N C ) is equal to the multiplication of the percent (P) by the maximum probable number (P M ) of classes may found in the satellite images, as follows: N C = P P M (26) Dataset Formatting and Storing deals with output centroid of K-Means algorithm. The image block corresponding or closest to centroid moment is stored in two dimensional dataset array (A), in which each block is converted into one dimensional vector to be one column in A. Such that, the width of A is the number of classes (N c ) while the height of A is equal to the number of pixels found in the block (i.e., B Max B Max ). 4.6 Classification Phase The classification phase is carried out after performing the training phase (enrollment). It can be achieved by using SVD method depends on the established dataset array A, where the i o, j o classification phase used the quadtree to segment the prepared image into non uniform blocks restricted between B Max and B Min. 2,2 2, 3 Then each square either leaved as it or subdivided into four Ds4 quadrants when it satisfies the partitioning conditions. Then, each block is assigned to the dataset array A to compute the 3, 2 3, 3 classification feature. According to this feature, the block is labeled with available classes. Since the SVD classification method needs to partition the image into predefine sized image (d) Compute the distance D block, quadtree partitioning method is used for segmenting the s4. image into addressed image blocks. Therefore, the Figure (2) The distance D S computation. implementation of quadtree partitioning method requires to set The implementation of K-Means depends on two input some parameters are related the partitioning conditions, which parameters, they are; the number of clusters (or classes) and the are used to control the process of partitioning. These control moment values of each block in the image. Actually, the parameters are given in the following: number of classes (N C ) in the prepared image is determined in 1. Maximum block size (B max ). the following steps: 2. Minimum block size (B min ). 1. Determine the number of pixels in satellite image N T by 3. Mean factor (β): represents the multiplication factor; when it is multiplied by global mean (M using the following relation: g ) it will define the value of the extended mean (M e ), i.e. M e =β M g. N T =W H (22) 4. Inclusion factor (α): represents the multiple factor, when it Where W represents the width of satellite image and H is multiplied by the global standard deviation ( ) it will represents the height of satellite image. define the value of the extended standard deviation ( ), 2. Determine the standard deviation ( ) to prepare image that i.e. e =α. is by employing equation (8) to be modified in the 5. Acceptance ratio (R): represents the ratio of the number of pixels whose values differ from the block mean by a following form: distance more than the expected extended standard = (23) deviation. The adopted SVD feature is estimated for each block to be Where is the mean of the prepared image that can be compared with that of the dataset A. This is first including the conversion of the block into one dimensional vector (V) and computed by the following relation: included in the dataset array A to be the sixth column, such that (24) the array will dimensioned as [(N c +1) (B Max B Max )].. The challenged problem is to fit the length of columns of the dataset array A with the vector V. This problem is over comes by down sampling the length of columns of A to be equal to the length of the vector V. The down sampling of each column elements is done by averaging process, in which the reducing ratio (R) is computed by dividing the length of current image block B L by the length of the A columns (i.e., B Max B Max ) as follows: (27) When the columns of the dataset array A are fitted, the SVD feature of current image block can be computed in comparison with dataset columns. The differences between the computed SVD are used to compute the similarity measure ( ) for the last column with that of its previous columns as follows: (28) Where, SVD k is the computed singular value decomposition feature of the k th class, and SVD k+1 is the singular value decomposition of the image block that need to be classified. The maximum value of refers to the class that image block is belonging to. The comparison leads to classify the image blocks.

6 RESULTS AND DISCUSSION The multiband satellite image used in the classification was capture by Landsat satellite, it cover the area of Baghdad city in Iraq. Figure (3) shows the six bands of used satellite image. The resolution of each band is 1024x1024 pixels, which carried acceptable range of informatic details about the image of consideration. One of the most important factors of using the Landsat Baghdad image is the different concepts of landcover appears in the image, which leads to different classes found in the image. Figure (4) Result of the image composition. Image transform is used to enhance the satellite image by using equation (12). It is applied on the three color components Band-1 Band-2 Band-3 (R, G, and B) of the image, which leads to converting the image from three bands into one band is better and suited for machine based analysis. The image preparation aimed to make the contrast of the considered image is full. Full contrast is achieved when choosing the values of Min 2 and Max 2 to be by using equation (13). Figure (5) shows the result of transformed and prepared image. Band-4 Band-5 Band-6 Figure (3) The used six bands of Baghdad city given by Landsat. The results of the dispersion coefficient (D) of used six bands are given in Table (1). It is shown that the greatest three values of the dispersion coefficients are belong to the bands (1, 2, and 3) respectively. Therefore, to compose these bands with each other for making one color image, it is assumed that the band (1) is stand for green (G), band (2) is stand for red (R), and band (3) is stand for blue (B) in the RGB colored image. Figure (4) shows the result of the composition process. Actually, the composed image enjoyed with more contrast and more visual details. Table (1) Resulted dispersion coefficient of the adopted six bands. Band D Figure (5) Results of image transform and preparation. 5.1 Enrollment Results The result of enrollment phase is a dataset stored in two dimensional array (A), the number of columns of this array is equal to the number of classes, while the number of rows of this array is equal to the length of the class. The length of the class is equal to the number of pixels contained in the image block, which can be determined by product the width by height of the block. The results of the uniform image partitioning is shown in Figure (6), in which the prepared image of resolution pixel is partitioned into image blocks each of size 8 8 pixel. The blocks greater than B Max lead to confuse the classification results, whereas the blocks less than B Min lead to

7 822 poor image parts and no information may found in image blocks. The moment of each image block was computed according to equation (21), the minimum and maximum resulted values of computed moment are shown in Figure (7). It is noticeable that the minimum value of the moment is zero, while the maximum value is The zero value refers to empty blocks, which have no information in, while the maximum value refers to much information found in that block. The application of the K-Means needs to set the range of expanding the clusters along the moment scale. Therefore, the range between the maximum and minimum values of the moment is , which is divided into five (N C= 5) of regions each of which extended by a maximum distance is equal to (D k = /5= unit). whereas Figure (9) displays the position of the image blocks that consisting in the dataset array A. It is observed that dataset had contained different classes, which confirms the correct path of clustering, where the resulted classes were far away from each other by an equivalent distances in the grey scale depending on the details of each class Water Resident with Resident without Open Land Figure (8) Behavior of five columns of five classes in the image. Class 5 Class 4 Figure (6) Result of uniform image partitioning (B Max =8 pixels). Class 3 Moments Blocks Figure (7) Sample range of resulted moment values. Finally, the dataset array A contains image blocks corresponding to the final centroids resulted from the application of the K-Means, each of these blocks represents a one column in the dataset array A sequentially. Figure (8) shows the behavior of these five columns that represent the labels of the discovered five classes of the image under consideration, Class Classification Results Figure (9) Resulted five classes. Class 1 In the SVD method, the finding of best values of control parameters and the best partitioning of the quadtree is very important problem since the control parameters govern the partitioning process that lead to intended classification. Figure (10) shows the best control parameters of quadtree partitioning method.

8 823 Maximum block size=8 Minimum block size=2 Acceptance ratio =0.2 Inclusion factor = 0.6 The classification result of the prepared image using the SVD method is displayed in Figure (11). It is shown that the distribution of classes along the image region was acceptable. The best values of control parameters make the partitioning process more accurate, which leads to accurate classification results. It seen that the results of image partitioning based on image homogeneity measurements are very acceptable. The result of the partitioning depends on the quantity of the uniformity for each block. The process of comparison was carried out pixel by pixel to guarantee the comparison result gave more realistic indication. The procedure is done by counting the number of pixels in the classified image that gave identify same class in the standard classified image. Then, the percent (P T ) of the identical classified pixels (C p ) to the total number of pixels (T p ) found in the image is computed as follows: Threshold =0.6 (29) Where, P T represents the overall accuracy (OA) of the proposed classification relative to the classification of the standard classified image given by GSC. Moreover, this relation can be employed to estimate the accuracy of each class in the image separately. This is carried out by examining pixels of classified image that identify same class in the standard classified image, which can be given in the following relation: Figure (10) Result of quadtree partitioning for control parameters. (30) Where, P k is the classification accuracy of k th class that represents the user's accuracy (UA), C c is the total number of pixels that classified as same as its corresponding pixels in the standard classified image given by GSC, and T k is the total number of pixels belong to the k th class in the classified image. Water Resident with Resident without Water Resident with Resident without Open Land Figure (11) Classified image using SVD method. Open Land To estimate the accuracy of the proposed two methods of satellite image classification, a standard image is classified by Geological Surveying Corporation (GSC) used for purpose of comparison. This standard image is classified by Maximum Likelihood Method using ArcGIS software version 9.3. The classification map in this image is shown in Figure (12), there are five distinct classes; there are five distinct classes; they are: water, vegetation, residential with vegetation (Resident -1), residential without vegetation (Resident -2), and open land, let we denote them as C 1 for class water and C 2 for class vegetation and C 3, C 4, C 5 for classes Resident with vegetation, Resident without vegetation, and Open Land respectively. Figure (12) The Standard satellite image classification given by GSC. Accordingly, the producer accuracy (PA) can be computed using the following relation: (31) Where, P p represents the producer accuracy (PA), and C p is the total number of pixels of each class in the standard classified image. The two parameters P c and P p are prepared to estimate both the commission error (E C ) and omission error (E O ) as follows: E C = 100- P K (32) E O = 100- P P (33) The use of equation (29) on the whole image gives best estimation for pixel classification rather than the use of random selected areas since the selection of small considered area may gave unstable result at each run of comparison due to the change of position of considered area. The evaluation results

9 824 SVD method is listed in Table (2), this table includes the overall accuracy and class accuracy for the SVD classification method. 100 Class Legend These results showed the SVD classification method was 80 C1 successful due to the percents of identical classes (P k ) were acceptable. It is noticeable that the class of Resident with 60 C2 (Resident -1) has less identical percent due to the details of such class is large enough to be described in the used image, while the class of water has a high identical percent due C3 C4 to it appeared in different spectral intensity in comparison with other classes, whereas; other classes are distributed moderately 0 C5 between the two mentioned classes. Also, it showed that the C1 C2 C3 C4 C5 overall accuracy of the classified satellite image is %, Classes while the total accuracy is about % when the Resident without (Resident -2) and Resident with vegetation (Resident -1) classes are regarded as same class. Figure (13) Figure (14) Producer accuracy of classes in SVD method. indicates that the class of Water showed high identification percent in comparison with that of standard image relative to other classes in the standard image. Moreover, Table (2) mentioned that the largest user's accuracy achieved with the Table (2) The results of SVD classification method. high accuracy for the class of water, the high value of user's accuracy has been found % for comparison between Standard Classified Image the results of the standard classified image and the SVD based C classified image, while the smallest user's accuracy was found 1 C2 C 3 C4 C 5 P K in the class of Resident with %. It is C concluded that the rest user's accuracy for the classes of satellite image are limited between the maximum and minimum percent C user's accuracy. On other hand, the high producer accuracy achieved for the class Resident without is C % and the smallest producer accuracy for the class Resident with is % the rest classes are C limited between the larger and smaller producer accuracy as shown in Figure (14), where the class of Resident with C has the smallest accuracy value. Also, Figure (15) describes the variation of each class in both user's accuracy and producer accuracy, where the user's accuracy classes: water, resident With, and Open Land class are greater than P p their producer accuracy, while the user's accuracy of and Resident Without are less than the producer accuracy of the standard classified image, which indicates the classes of water, Resident with vegetation and, are more changed compared with the other classes. OA Classified Image Producer Accuracy Classification Score C1 C2 C3 C4 C5 Classified Image Class Legend C1 C2 C3 C4 C5 Percentage of Classes C1 C2 C3 C4 C5 Classes User's Accuracy Producer Accuracy Figure (13) Classes accuracy in SVD method. Figure (15) Relation between producer and user's accuracy of classes By using SVD Method..

10 CONCLUSIONS In this paper, the overall accuracy of the classified satellite image is %, which can be rised to be about % when regarding both Resident without vegetation and Resident with vegetation classes as same class. Where the used of quadtree serves the classification stage due to the block size was smaller time by time till reaching to spectrally homogenous region. And, the classification results of SVD method show that the variation of each class in both user's accuracy and producer accuracy, where the user's accuracy classes: water, resident With, and Open Land class are greater than their producer accuracy, while the user's accuracy of and Resident without vegetation are less than the producer accuracy of the standard classified image, which indicates the classes of water, Resident with vegetation and, are more changed compared with the other classes. 7. REFERENCES [9] Sathya, P., and Malathi, L., "Classification and Segmentation in Satellite Imagery Using Back Propagation Algorithm of ANN and K-Means Algorithm", International Journal of Machine Learning and Computing, vol. 1, no. 4, October [10] Rowayda, A. S.," SVD Based Image Processing Applications: State of The Art, Contributions and Research Challenges", International Journal of Advanced Computer Science and Applications, Vol.3, P [11] Bjorn F., Eric B., IreneWalde, Soren H., Christiane S., and Joachim D., 2013, "Land Cover Classification of Satellite Images Using Contextual Information", ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume II-3/W1, [12] Ankayarkanni and Ezil S. L., "A Technique for Classification of High Resolution Satellite Images Using Object-Based Segmentation", Journal of Theoretical and Applied Information Technology, Vol. 68, No.2, and ISSN: , [13] Harikrishnan.R, and S. Poongodi, "Satellite Image Classification Based on Fuzzy with Cellular Automata", International Journal of Electronics and Communication Engineering (SSRG-IJECE), ISSN: , volume 2 Issue 3, March [14] Akkacha B., Abdelhafid B., and Fethi T. B., "Multi Spectral Satellite Image Ensembles Classification Combining k-means, LVQ and SVM Classification Techniques", Indian Society of Remote Sensing, [15] Thwe Z. P., Aung S. K., Hla M. T., " Classification [1] Chijioke, G. E. " Satellite Remote Sensing Technology in Spatial Modeling Process: Technique and Procedures", International Journal of Science and Technology, Vol. 2, No.5, P , May 2012,. [2] Baboo, Capt. Dr.S S., and Thirunavukkarasu,"Image Segmentation using High Resolution Multispectral Satellite Imagery implemented by FCM Clustering Techniques", IJCSI International Journal of Computer Science Issues, ISSN (Print): ISSN (Online): , vol. 11, Issue 3, no 1, S., May [3] Sunitha A., and Suresh B. G., " Satellite Image of Cluster Area For satellite Image", International Classification Methods and Techniques: A Review", Journal of Scientific & Technology Research Volume 4, International Journal of Computer Applications, pp. Issue 06, 2015, , Volume 119 No.8, June [16] Anand, U.; Santosh, K. S. and Vipin, G.S., " Impact [4] Mayank T., "Satellite Image Classification Using of features on classification accuracy of IRS LISS-III Neural Networks", International Conference: Sciences images using artificial neural network ", International of Electronic Technologies of Information and Journal of Application or Innovation in Engineering and Telecommunications, Management,V. 3,P , October [5] Al-Ani L. A. and Al-Taei M. S.,"Multi-Band Image [17] Neil M., Lourenc M., and Herbst B. M.," Singular Classification Using Klt and Fractal Classifier ", Value Decomposition, Eigenfaces, and 3D Journal of Al-Nahrain University Vol.14 (1), pp.171- Reconstructions", Society for Industrial and Applied 178. Mathematics, Vol. 46, No. 3, pp , [6] Hameed M. A. Taghreed A. H., and Amaal J. H., [18] Brindha S., "Satellite Image Enhancement Using 2011, " Satellite Images Unsupervised Classification DWT SVD and Segmentation Using MRR MRF Using Two Methods Fast Otsu and K-means ", Model", Journal of Network Communications and Baghdad Science Journal, Vol.8 (2), Emerging Technologies (JNCET), Volume 1, Issue 1, [7] Bin, T.,Azimi-Sadjadi, M. R. Haar,T. H. V.Reinke, D., " Neural network-based cloud classification on [19] Ranjith K. J., Thomas H. A., and Mark Stamp, satellite imagery using textural features", IEEE, Vol. "Singular value decomposition and metamorphic 3, p , detection", Springer-Verlag France, J Comput Virol [8] Márcio L. Gonçalves1, Márcio L.A. Netto, and José Hack Tech, A.F. Costa, "A Three-Stage Approach Based on the [20] Ilya S., James M., George D., and Geoffrey H.," On Self-organizing Map for Satellite Image the importance of initialization and momentum in Classification", Springer-Verlag Berlin Heidelberg, pp. deep learning", International Conference on Ma-Chine , Learning, Atlanta, Georgia, USA, volume 28, 2013.